In the context of intense global competition for dominance in emerging technologies, artificial intelligence (AI) has become one of the most strategic, contested, and sensitive areas of the 21st century. This struggle, mainly led by China and the United States, is not limited to the development of sophisticated tools but also reflects an ideological and geopolitical confrontation involving democratic values, state control, intellectual property, and technological hegemony.
Artificial intelligence, as a driver of transformation in the economy, defense, and society in general, has become a symbolic and real battlefield for both powers. In this scenario, the Chinese startup DeepSeek has emerged strongly, presenting advances that seem to rival the most advanced language models developed in Silicon Valley. The company, based in Hangzhou, has claimed that its language models — DeepSeek-V3 and R1 — not only rival those of leading companies such as OpenAI and Meta but do so at a fraction of the cost.
These advances by DeepSeek have quickly attracted the attention of developers, investors, and governments. However, a series of recent investigations, including exclusive revelations by Reuters, have placed DeepSeek at the center of an international controversy. The accusations against DeepSeek range from opaque business practices to alleged direct links with the Chinese military and maneuvers to evade technological sanctions imposed by the United States. Below, ITD Consulting details the new DeepSeek controversy.
The meteoric rise of DeepSeek: Innovation, efficiency, and doubts
DeepSeek became internationally known in 2024 when its developers claimed to have trained a large language model with a surprisingly low computational investment, estimated at just 5.58 million dollars. In an ecosystem where training models like GPT-4 can cost hundreds of millions of dollars, this efficiency raised both enthusiasm and skepticism. Many experts considered it practically impossible to reach comparable levels without access to state-of-the-art hardware and computational resources.
What most unsettled analysts was the lack of transparency in DeepSeek’s technical and scientific processes. While companies like OpenAI, Anthropic, or Google DeepMind publish papers, benchmarks, and architectural descriptions of their models, DeepSeek opted for a more ambiguous communication strategy. Few details have been offered about the exact architecture of DeepSeek’s models, the datasets used, the training time, or the human and computational resources involved. This secrecy not only hinders validating their achievements but also raises suspicions about the possibility that their results depend on external or undisclosed sources.
Despite this, DeepSeek’s proposal has resonated strongly in the Asian market and industrial sectors seeking efficient and economical AI models. DeepSeek’s ability to offer fast responses, automatic translations, text generation, and natural language processing at low cost has driven its adoption in areas as diverse as education, medicine, robotics, and engineering.
Moreover, DeepSeek has leveraged the Chinese technological ecosystem, which combines state funding, support from technology incubators, and a growing talent base in artificial intelligence. This environment has facilitated the rapid scaling of DeepSeek and the consolidation of its position as a global competitor. Nevertheless, this success also fits within a broader national strategy to position China as a world leader in AI by 2030, a plan that includes large public and private investments, infrastructure development, and adapted regulation.
Technological evasion: Avoiding sanctions through geopolitical engineering
One of the most delicate aspects of this case involves accusations from the U.S. government, according to which DeepSeek participated in a complex evasion strategy to acquire restricted technology. Since 2022, the Joe Biden administration has tightened restrictions on the export of advanced chips to China, especially those designed for artificial intelligence applications. Among these are Nvidia’s H100 chips, essential for training complex models.
According to the Reuters-cited investigation, DeepSeek allegedly used shell companies in Southeast Asian countries — such as Malaysia, Vietnam, and Singapore — to indirectly access these chips. DeepSeek’s method consisted of renting computing capacity in data centers located outside China, thus avoiding direct acquisition of sanctioned hardware. Although technically this would not be a direct violation, this kind of maneuver represents a breach of the spirit of the sanctions and reflects the difficulty of current regulations to control virtual and globalized infrastructures.
This case has also served as an example of the structural limitations faced by current regulatory frameworks. In a world where cloud computing allows virtual infrastructures to operate from anywhere on the planet, controls relying exclusively on physical borders or tangible hardware fall short against the distributed capabilities of companies like DeepSeek. This situation has put Western authorities in check, forcing them to rethink policies to contain the transfer of strategic technologies without hindering innovation and legitimate trade.
The complexity of this maneuver also reflects the sophistication of China’s technological strategy, which combines state and private resources to circumvent restrictions and maintain an accelerated pace in the race for artificial intelligence. This approach includes not only creating networks of shell companies but also cooperating with international partners and using global cloud platforms to distribute and train models.
Furthermore, the lack of a unified response from Western countries has complicated the effective enforcement of sanctions, showing that digital technologies and their infrastructure require much more agile and coordinated regulatory frameworks and international cooperation mechanisms than those used to control traditional physical goods.
Military connections: Artificial intelligence as a dual-use weapon
Perhaps the most serious aspect of the accusations relates to the links between DeepSeek and the People’s Liberation Army (PLA). According to documents cited by State Department officials, the company DeepSeek has been mentioned more than 150 times in procurement records of entities linked to Chinese defense. These mentions of DeepSeek include collaborations with military universities, tactical simulation centers, and state-owned companies that develop autonomous systems for military use.
A concrete example of these collaborations is the use of DeepSeek models by a university in northern China which, in a documented experiment, generated more than 10,000 tactical scenarios in less than a minute—a task that would have taken 48 hours with human resources. Likewise, the firm Chongqing Landship, specialized in military robotics, integrated a DeepSeek AI system into an autonomous vehicle presented at a defense exhibition in Abu Dhabi.
These developments show how AI can be easily used for dual purposes: civilian applications such as automatic translation or medical assistance can be repurposed for military analysis, drone control, or cyber warfare. Given the lack of separation between the civil and military spheres in the Chinese technological governance model, these links generate great concern among Western intelligence agencies, which fear that DeepSeek is part of a state strategy to rapidly advance autonomous military capabilities.
Likewise, this duality of DeepSeek has been a central topic in international debates on AI ethics, where transparency in the technological development chain becomes key to preventing artificial intelligence from becoming a multiplier of armed conflicts or a tool for mass surveillance and political repression.
Surveillance and privacy: Can data be a weapon of power?
Another key axis of this controversy is how DeepSeek handles its users’ data. DeepSeek’s models are available on platforms such as Amazon Web Services, Google Cloud, and Microsoft Azure, allowing millions of users outside China to access its services. However, a report from the Stanford Cyber Policy Center warns that DeepSeek collects an unusually large amount of data, including attachments, keystroke logs, voice patterns, user behavior, and sensitive personal data.
This raises a crucial dilemma: under Chinese cybersecurity laws, all technology companies must cooperate with authorities if requested for information. This means any data collected by DeepSeek could end up in the hands of the Chinese government, including private conversations, strategic documents, or internal analyses conducted by companies or citizens from other countries.
The contrast between this threat and the accessibility of DeepSeek models on American cloud services has been the subject of harsh criticism. Although agencies such as the Pentagon, NASA, and Congress have formally banned its use, public versions of DeepSeek remain available without restrictions for individual users or companies, revealing a dangerous disconnect between public policy and commercial practice.
Furthermore, the growing global dependence on artificial intelligence services like DeepSeek poses an even greater challenge for privacy and information security. Increasingly, AI infrastructures are integrated into critical services—finance, health, education—increasing the risk of data leaks or malicious manipulation.
Technological governance: Toward a new regulatory framework
Digital policy experts such as Sanchit Vir Gogia have pointed out that the DeepSeek case represents a clear warning about the limits of the current technological control approach. In a hyperconnected and virtualized environment, it is not enough to regulate what type of hardware can cross borders. It is necessary to supervise how services are used, who controls them, and for what purpose. To achieve this, a combination of:
- Algorithmic behavior audits
- Certifications of technological neutrality
- International supervision of data flows
Clear rules on the use of AI by companies linked to authoritarian governments
This kind of approach, according to Gogia, can only arise from strong international cooperation, involving both technological democracies and multilateral organizations. Without this shared framework, AI development will be at the mercy of powers that can use it as a tool for internal control or geopolitical influence.
In addition, there is a highlighted need for the business community to take an active role in defending ethical and responsible artificial intelligence development. This includes adopting independent audits, clear data management policies, and a commitment to transparency.
For now, DeepSeek has not been formally included on the U.S. Department of Commerce’s list of sanctioned entities. However, U.S. officials have hinted that actions in this regard are being considered. If sanctions were ultimately imposed, DeepSeek would face severe restrictions on access to hardware, software, and clients outside China, which could slow its international projection and affect other technological ventures in the country.
Moreover, it is speculated that the U.S. could tighten access to public cloud platforms for companies linked to authoritarian regimes, a measure that would profoundly change how large-scale AI models are trained and deployed. This would have effects not only in China but also on startups and universities worldwide, which would need to demonstrate their neutrality and traceability before accessing sensitive computational resources.
On the other hand, the DeepSeek case could incentivize other countries to accelerate their own autonomous technological development policies, increasing fragmentation of the global AI ecosystem and making international cooperation even more difficult.
The DeepSeek case is much more than a commercial or technological dispute. It is a clear example of how artificial intelligence development has become a space of struggle for power, digital sovereignty, and ethics in the 21st century. On one hand, the company DeepSeek represents a milestone for China, which manages to position itself at the level of the great technological powers. On the other, it embodies the inherent risks of lack of transparency, military use, and personal data exploitation.
For the international community, the challenge is twofold. Open and collaborative innovation must be encouraged, but at the same time, it must be ensured that these tools are not used as instruments of surveillance, coercion, or geopolitical domination. Achieving this balance will require new alliances, a profound rethinking of digital governance, and a firm defense of democratic values in the age of artificial intelligence.
DeepSeek has shown that China can compete at the highest level. Now the world must decide how to respond: with fear, fair competition, or vigilant cooperation. Only then will it be possible to harness AI’s transformative potential without sacrificing security and fundamental freedoms. If you want to learn more about the DeepSeek case, AI regulations, and how this may impact your operations, write to us at [email protected]. We have a team of experts ready to advise you and keep you at the forefront of technology and artificial intelligence.
EN 
ES